Home

The design of ScanSpec - a scanning imaging FTIR

Contents

1. MCT detector a 85 mm a 125 mm a 3 125 mm a 4 90 mm InSb detector a 92 mm a 125 mm a 3 120 mm a s 93 mm Result MCT detector with aperture fully open and distance d a 59 1 m FOV 0 209 InSb detector with aperture fully open and distance d a 59 1 m FOV 0 210 59 Appendix 1 Measurement 2 Parameters Same as for measurement except distance d a 88 2 m Measurements To obtain the corresponding angles a a2 a3 az multiply by 6 50 10 mm InSb detector a 145 mm a 195 mm a 3 190 mm a s 140 mm Result InSb detector with fully open aperture and distance d a 88 2 m FOV 0 220 Measurement 3 Parameters Same as for measurement except distance d a 40 1 m aperture of the blackbody 6 2 mm Measurements To obtain the corresponding angles a a2 a3 a4 multiply by 1 429 10 mm InSb detector a 62 mm a 81 mm a 3 74 mm a s 158 mm Result InSb detector with fully open aperture and distance d a 59 1 m FOV 0 198 Measurement 4 Parameters aperture B 3 2 mm aperture AB 6 4 mm aperture of the blackbody 12 5 mm blackbody temperature 1209 C distance d a 59 1 m Measurements To obtain the corresponding angles a a2 a3 az multiply by 9 70 10 mm InSb detector ai 71 mm a 2 63 mm a s 60 mm a 4 73 mm Result InSb detector with aperture 3 2 mm and distance d a 59 1 m FOV 0 130 Measure
2. chapter 8 there are examples of what this GUI can look like There are eleven standard scan programs defined in ASCII text files The most important parameters in those files and the ones needed to understand the signal processing are the following 36 e Image size The spatial size of the spectral image can be chosen to be 8x8 16x16 32x32 or 64x64 pixels e Number of interferograms that should be added for the acquisition of one spectrum The more interferograms added per spectrum the higher is the signal to noise ratio e Spatial oversampling If the spatial oversampling is N N spectra are acquired per pixel field of view per 0 10 when the mirror is turned around the z axis e Spectral resolution The spectral resolution can be chosen to 1 2 4 8 and 16 cm The standard scan programs uses 4 and 16 cm The lower resolution the more spectra per second can be acquired 7 2 Signal processing The positions of the rotary tables sampled by the MM with a certain frequency are first saved in the memory of the MM After the acquisition is finished the data is transfered to the FTIR computer via a serial connection The maximum number of samples is limited to 4000 by the MM Also it takes a long time to read the positions recorded by the MM when the scanning of the mirror is finished One reason why it takes so long to send the data from the MM is that the positions can for some strange reason only be sent to the FTIR compute
3. spatial FTS For spatial FTS one axis on the 2D IR detector array is used to sample the interferogram while the other axis is used to acquire spatial information in one dimension The other spatial dimension has to be sampled in time There are several different optical designs for imaging spectrometers based on spatial FTS see Sellar One example of a possible design will be described below A spectrometer using a grating to achieve path differences This type of spectrometer is called a digital array scanned interferometer DASI It is based on a Michelson interferometer using the principle shown in Figure 5 but it has no moving parts A grating is used not to disperse light but to achive the optical path difference necessary for the interferogram to be recorded This can be seen in Figure 6 The interferograms and one spatial dimension can be recorded simultaneously by the 2D detector array The other spatial dimension necessary to form a spectral image has to be scanned with a mirror or by moving the whole spectrometer mirror light from i aperture and grating collimator i light to detector array Figure 6 An interferomer where a grating is used to produce the optical path difference necessary to produce the interferogram The 2D detector array records interferograms in one direction and a spatial dimension in the other direction According to Hammer a DAST has the potential of achieving hi
4. Sweden Before we will learn more about spectral images in the next chapter I would like to take the opportunity to thank all the very helpful people at FOA in Link ping who have contributed to this project especially P r Nilsson and Claes Nelsson 2 Spectral images and their military use 2 1 What is a spectral image People view the world in color and draw conclusions about objects due to the color of the object Human eyes or a normal color video camera produce spectral images defined below with very low spectral resolution only three frequency bands It is easier to recognise objects on a color TV compared to on a black and white TV that is objects are easier to detect if the light from many frequency bands is recorded Compared to a color video camera or human eyes with only three frequency bands an imaging spectrometer can increase the number of recorded frequency bands to thousands or more The human eye covers less than 10 per cent of the spectral region available to imaging spectrometers Breckinridge 1996 Imaging spectrometers will provide us with further information in many fields of research ranging from studying outer space to looking at microscopic objects The definition of what a spectral image is can vary slightly My definition of a spectral image is a three dimensional data entity that describes the spectral radiance of a scene as a function of two spatial variables and one spectral variable The two spatial v
5. The imaging spectrometer from Pacific Advanced Technology Their instrument is shown in Figure 4 More information can be found at their web site internet Patinc Hinnrichs ImSpector N17 The Finnish company Specim Spectral Imaging Ltd sells a spectral line imager based on dispersion See internet Specim The N17 costs 4700 USD and 19000 USD together with a suitable InGaAs IR camera email Parameter 3 4 Imaging spectrometers based on Fourier transform methods Fourier transform spectrometry FTS For readers not familiar with Fourier transform spectrometry I will briefly explain the main idea mirror mirror detector Figure 5 The basic principle of Fourier transform spectrometry The basic principle of Fourier transform spectrometry FTS is shown in Figure 5 The optical layout can vary but the principle is the same Light enters from the left in the figure a beam splitter 15 reflects 50 per cent of the light towards the top mirror and the rest of the light hits the mirror to the right The light is reflected on the mirrors and the light from the two optical paths are reunited at the beam splitter The light from the vertical and the horizontal optical paths interfere with each other and hit the infrared detector or a two dimensional array of IR detectors By moving the right mirror and simultaineously studying the interference pattern of the light at the detector an interfer
6. FOA Department of IR systems Link ping Sweden The CD includes all the programs written for ScanSpec the data from the first measurements discussed in this report and the complete documentation of the system including User s manual Software manual the FOA report manuals and this report Email Coulson email sent from Paul Coulson p coulson dial pipex com Bomem Inc to Frans Lundberg FOA 17 June 1998 Hinnrichs email sent from Michele Hinnrichs micheleh syv com Pacific Advanced Technology to Frans Lundberg FOA 12 August 1998 Parameter email sent from Ched Kent ck parameter se Parameter AB to Frans Lundberg FOA 25 August 1998 50 Fax Bernard Halle 980710 fax from W Vosskuhler Bernard Halle Berlin Germany to me 10 Juli 1998 Spectrogon 980625 fax from Curt Erik Lundquist Spectrogon T by Sweden to me 25 June 1998 Internet Aerotech home page of Aerotech Inc www aerotechinc com June 1998 Anorad home page of Anorad www anorad com June 1998 Ball internet address of Ball Aerospace amp Technologies Corp www ball com aerospace pt html June 1998 Daedal internat address of Parker Hannifin Corporation Daedal Division www daedalpositioning com August 1998 Patinc internet address of the company Pacific Advanced Technology Inc in California USA www patinc com August 1998 Rockwell internet address for infrared focal plane arrays from Rockwell and Boeing www rsc rockwell
7. No 1 Z O 1 40 lt 20 T 20 40 80 100 120 3 20 40 60 80 100 120 O 0 i 2 a 4 5 6 2 Oo 2 4 6 8 mas Maes wna es o 16972 ej af 4 78376 a EI 02 Eo 11186 4f E OO 42585 a gt Figure 28 This figure shows the Volvo after the window was opened In the right image only the exhaust emissions can be seen 45 9 Results and possible improvements 9 1 Distortion The ScanSpec imaging system is not free from image distortion but when the angle 2 is kept close to zero the distortion will be low It is recommended to use 2 angles less than 25 If that recommendation is followed there should be no problems with distortion unless very low distortion is needed The distortion is known exactly and depends only on the start 02 and gt angles therefore the distortion could be corrected for Most of the distortion could be avoided with some changes in the software that governs the scan pattern This issue is briefly discussed in the Software manual section Improvements and changes CD ScanSpec 98 1 9 2 Are the goals of the project met The main goal of the ScanSpec project was to make an imaging spectrometer from an existing non imaging spectrometer and that the existing spectrometer would limit the performance of the total system using the resources available The Bomem spectrometer is limiting the total performance in the most important parameters such as spectral resolution spatial r
8. The two electrical information paths marked with thick arrows were made by me 34 The other thick arrow that goes between the MM and the FTIR computer demands some explanation The wiring which is shown in Figure 21 For a definition of the pins on the MM and the PIO12 se MM and PIO FTIR computer Parallel port PIOI2 serial port pin18 pin37 pin21 pin15 ground PAO ground This part of the wiring is placed in the connector at the GPIO port of the MM pinl2 pinl3 pin3 output 1 output 2 5V G P I O port MM Figure 21 The wiring between the FTIR computer and the MM motion controller The connection between the serial port of the FTIR computer and RS232 port of the MM is used to send commands or whole motion control programs to the MM from the FTIR computer The FTIR computer receives information about recorded positions from the MM over the same wire The wire from output 1 of the MM to the pin5 of the parallel port of the FTIR computer is used to send a trigger pulse from the MM to the FTIR computer to synchronise the events between the FTIR computer and the MM The parallel port of the FTIR computer is read by a C program written by me The wire from output 2 of the MM to pin37 of the PIO12 acquisition board of the FTIR computer is used to send a trigger signal from the MM to the FTIR computer when the acquisition of spectra should start The PIO12 board is read by the Acquire software 35 7 The ScanS
9. a specific wavelength A solar telescope was placed so that the focal plane of the telescope coincided with the entrance slit of the spectrometer The light that passed through the spectrometer was recorded on a photographic plate By moving the slit with a constant velocity over the disc of the sun and keeping the photographic plate at the same position Hale was able to record the intensity of the light for a specific wavelength in one spatial dimension To record the intensity for another wavelength he had to adjust the spectrometer and make another scan Breckinridge The idea of recording both spatial and spectral variations of radiation has been around for a century or so However it is only the last decade or two that complete systems that can record a spectral image with two spatial dimensions and one spectral dimension have been built Not until the last few years have imaging spectrometers been available commercially for the MWIR and the LWIR regions and the market seems to be limited to only a few products I believe that the development of new imaging spectrometers for the MWIR and the LWIR regions will be rapid the coming few years due to the increased size and performance of 2D arrays of IR detectors See internet Rockwell and internet Sensors for commercially available 2D arrays of IR detectors 2 3 The military use of infrared spectral images Infrared spectral images are important in military research to measure optical sign
10. blur has shown to be a minor problem According to measurements see Appendix 1 the field of view of the telescope is approximately circular with a diameter of 0 20 when the largest apertures diameter 6 4 mm are used Assume that we would like to scan the mirror so that one spectra will be recorded for every 0 20 Since the spectrometer acquires 65 spectra s in the fastest mode the mirror has to turn with the speed of 13 s if the scanning is done by turning the mirror around the 0z rotation axis If the scanning is done around the the rotation axis only half of that speed is needed 6 5 s The minimum speed needed for the target following feature is 20 s for the O rotation axis and 10 s for the 2 rotation axis Those speeds determines how fast the tables should be The needed acceleration of the tables is another important issue The goal that the spectrometer should limit the performance of ScanSpec means in this case that the time when the spectrometer is not acquiring useful information the time while the mirror is not moving with a constant velocity to scan a line in the spectral image should be small compared to the total acquisition time One question that arises when deciding on the scan pattern is what the spatial sampling frequency should be The angular sensitivity of the InSb detector and the MCT detector together with the narrow angle telescope has been measured see appendix 1 The field of view of the telescope set
11. com mct_fpa December 1998 Sagebrush internet address of Sagebrush Technology Inc www sagebrushtech com August 1998 Sensors internet address of Sensors Unlimited Inc www sensorsinc com December 1998 Specim internet address of the company Specim Spectral Imaging Ltd in Finland www specim fi August 1998 Webster electronic dictionary of the English language based on Merriam Webster s Collegiate R Dictionary Tenth Edition Available at www facstaff bucknell edu rbeard diction html October 1998 Phone Melles Griot 980626 phone call from Urban Conradsson Melles Griot in Sweden to me 26 June 1998 Spectrogon 980618 phone call from Kurt Erik Lundquist Spectrogon T by Sweden to me 18 June 1998 Edmund Scientific 980624 phone call to Edmund Scientific 1 609 573 6250 24 June 1998 Presentation Villemaire Andr Villemaire Bomem Inc Jean Giroux Bomem Inc and Ron J Rapp Air Force Wright Laboratories An Imaging Fourier transform Spectrometer Third workshop on Infrared Emission Measurents by FTIR Quebec Canada February 1996 51 Appendix 1 Appendix 1 Measurements of field of view angular sensitivity and focusing range Introduction This text desribes measurements performed in September 1998 with the Bomem MR254 spectrometer together with the narrow angle telescope SMY 02 The sensitivity of the InSb and the MCT detectors has been studied for different angles off the optic
12. detectors simultaneously The spectrometer has a built in cold black body source used to optically subtract the internal radiation of the optics in the MR254 Both of the detectors and the black body source are cooled with liquid nitrogen The MCT detector is called Detector A and the InSb detector is Detector B The spectrometer has three apertures one between the telescope and the spectrometer and one in front of each of the two detectors There are seven settings of the diameter of the apertures 6 4 4 5 3 2 2 2 1 6 1 1 and 0 8 mm ScanSpec is currently designed to be used with all apertures set to 6 4 mm but it can be used with other aperture settings The MR254 in the ScanSpec system uses a narrow angle telescope also from Bomem Bomem specifies the field of view of the telescope to 0 27 but measurements at a distance of 60 to 90 m show that the field of view is closer to 0 20 see appendix 1 The spectrometer is controlled by a PC computer called the FTIR computer and a program developed by Bomem called Acquire While the spectrometer is collecting interferograms the data is sent from the spectrometer to an acquisition board in the FTIR computer via an optical fiber The memory of the acquisition board is limited to 64 Mbyte which is the maximum amount of data that can be collected by the spectrometer Each point of each interferogram is stored with 2 bytes so a maximum of approximately 32 million interferogram points can be ac
13. effective in detecting targets A number of target background combinations of military interest in forest and desert environments were investigated using data from a non imaging spectral database Eismann3 Eismann4 and Schwartz2 in Eismann The result was that a small number of spectral bands can provide good target detection performance if the bands are chosen appropriately Further the optimal spectral bands tend to be in the region from 8 5 um to 11 um or even 10 5 um The database was also used to estimate the minimal spectral resolution needed for an imaging spectrometer to detect military targets in the IR region The result was that a resolution of 250 nm is acceptable but 100 nm can be set as a goal for an imaging spectrometer in order to be sure that the spectral resolution of the instrument will be adequate Of course the optimal frequency region for target detection in Swedish environments could be somewhat different from the one mentioned above Eismann When discussing the spectral resolution that is needed to detect targets we will divide infrared spectral sensors into three groups multi spectral hyper spectral and ultra spectral sensors according to Breckinridge 10 Multi spectral instruments able to record only a few spectral bands are called multi spectral For example the human eye is a multi spectral sensor A spectral resolution of AMA 0 1 can be used for terrain classification Hyper spectral instruments A spe
14. optics SPIE 1997 Holst Gerald C Holst Electro optical Imaging System Performance Winter Park FL USA JCD Publishing 1995 Manuals Frans Lundberg User s manual and Software manual for ScanSpec internal FOA report FOA D 98 00414 615 SE December 1998 MM The User s manual of the motion controller MM4005 from Newport Inc PIO The documentation of the aquisition board PIO 12 from Keithley Pedrotti Frank L Pedrotti and Leno S Pedrotti Introduction to optics 2nd edition Prentice Hall 1993 Physics Carl Nordling and Jonny Osterman Physics Handbook 4th edition Lund Sweden Studentlitteratur 1987 Schwartz2 C R Schwartz et al Thermal Multispectral Detection of Military Vehicles in Vegetated and Desert Environments Signal and Data Processing of Small Targets Proceedings of the SPIE vol 2742 Orlando FL USA 2742 30 April 1996 Sellar R Glenn Sellar and J Fruce Rafert Fourier transform imaging spectrometer with a singel toroidal optic Applied Optics vol 34 no 16 June 1995 Villemaire A Villemaire S Fortin J Giroux T Smithson R Oermann An Imaging Fourier Transform Spectrometer SPIE vol 2480 1995 Watson Edward A Watson Robert A Muse and Fred P Blommel Aliasing and blurring in microscanned imagery SPIE vol 1689 1992 CD ScanSpec98 1 The software of ScanSpec version 98 1 First version is available on a compact disc from the
15. optimistic and their testing must have been done under very optimum conditions Those figures cannot be used testing is necessary The tests show that the used mirror mounts holds at least 300 N each which is well enough for the application This figure corresponds to a stress of 4 MPa 4 N mm The producer of the Araldite 2012 stated a figure of 15 MPa 29 6 3 The mirror turning system This section will serve as a brief description of why the chosen mirror turning system was favored and also serve as some advice for someone who intends to buy similar equipment The most important parameters for the mirror turning system follows e Speed At least 10 s for dx dt and 20 s for dy dt provide for the target following feature e Acceleration The faster the better An acceleration of 100 s is adequate for ScanSpec e Angular precision The field of view of one pixel in the spectral image is 0 10 We want the maximum error of the angles 2 and to be 20 of a pixel 0 02 Then the maximum error of the angle x must be 0 02 2 0 01 and the maximum error of y must be 0 02 e Control It must be possible to easily control the mirror turning system from a PC e Torque The torque must be high enough to turn the mirror so quickly that the turn around time for the mirror is low compared to the effective scan time e Range The system must cover the range of and O as shown in Figure 8 page 21 e Delivery time As soon
16. the uncertainty in the determination of Ad which affects the results when the sensitivity changes with Ao rapidly The error of Ad is estimated to 0 005 58 Appendix 1 Measurement 3 The field of view for different aperture settings The definition of field of view and how it was estimated Let us assume that the contour of the 0 5 sensitivity as in figure 2 and 3 is almost circular This assumption is good for the InSb detector but a poor one for the MCT detector Four angles were measured each time the field of view was to be determined These angles a a2 a3 a4 are depicted in Figure 9 Since the center of the field of view do not necessarily have to be exactly the same as the center of the visual field of view of the CCD camera the field of view was estimated the following way The position of the center of the field of view was estimated by aj a2 2 a3 a4 2 in the coordinate system in Figure 9 Then half of the field of view is estimated as the average of the four distances from the center of the field of view to the end of the arrows AM gt Aro Figure 9 How the field of view was estimated The measurements Measurement 1 Parameters aperture A fully open aperture B fully open aperture AB 6 4 mm aperture of the blackbody 12 5 mm blackbody temperature 1212 C distance d a 59 1 m Measurements To obtain the corresponding angles a a2 a3 az multiply by 9 70 10 mm
17. the cost of coating the glass The choice of surface coating material is less important for the price Alternatives Three different alternatives for the mirror were thoroughly investigated 1 A gold coated mirror with a high surface flatness 0 75 um Spectrogon could deliver such a mirror in 10 to 12 weeks for between 60 000 and 70 000 SEK phone Spectrogon 980618 Also Melles Griot could make such a mirror for around 65 000 SEK phone Melles Griot 980626 Such a mirror would be between 30 and 40 mm thick and weigh 10 to 13 kg with the size specified above 2 A thinner mirror with a medium flatness A flatness of around 9 um can be obtained with normal float glass about 15 mm thick Spectrogon could deliver a 15 mm aluminum coated mirror for 28 600 SEK in about 6 weeks after order fax Spectrogon 980625 Bernard Halle could sell an aluminum coated mirror with the mentioned flatness and thickness for 2100 DEM fax Bernard Halle 980710 which is a much better deal The weight of a mirror with the given dimensions is about 5 kg 3 A thin off the shelf mirror Edmund Scientific New Jersey USA sells a 408 x 609 x 6 mm aluminum coated float glass mirror for 200 USD The flatness is 6 to 8 um according to phone Edmund Scientific 980624 but these figures are not guaranteed by Edmund Scientific The mirror is too large but since it is so thin it can be cut with an ordinary glass cutter The weight after it is cut to the desired s
18. this distance is supposed to be 30 m Measurement 2 Angular sensitivity These were the parameters for the measurements Aperture A fully open Aperture B fully open Aperture AB 6 4 mm aperture of the blackbody 12 5 mm blackbody temperature 1212 C distance d a according to Figure 1 59 1 m The radiation was measured for different angles of Ad and A0 The angle Ad is shown in Figure 1 This angle is positive if the blackbody aperture is on the right hand side of the optical axis of the telescope AB is the elevation angle If the blackbody aperture is above the optical axis of the telescope AQ is positive The measurements were performed for both detectors at every crossing of a 20 x 20 mm grid on the cardboard in front of the blackbody This corresponds to measurements at every 0 019 for the angles Ad and AB 13 x 13 values were acquired for both detectors The background level was subtracted from the result The data were then normalized so the maximum value is 1 00 The 53 Appendix 1 absolut values are of no interest for these measurements The results for the MCT and the InSb detector can be seen as contour plots in Figure 2 and 3 0 15 0 1 0 05 AO 0 deg 0 05 0 1 0 1 0 05 0 0 05 0 1 0 15 A9 deg Figure 2 The normalized sensitivity of the spectrometer within the field of view The narrow angle telescope and the MCT detector were used Figure 2 shows the sensitivity of the MCT
19. was able to acquire valuable data The next version DREV ISR I is an improved field instrument Eglin ISR is a newer 1996 instrument also based on a Bomem FTIR spectrometer It is a compact prototype for airborne use These are some of the specifications for the Eglin ISR Spectral range 3 5 5 um compatible to LWIR also Spectral resolution 1 2 4 8 16 32 64 128 cm computer selectable i Spatial resolution 8x8 pixels Frame rate 60 frames s at 4 cm resolution Maximum recording time 15 s at 4 cm resolution My conclusions of the Eglin are that it is very fast and still has a high spectral resolution The speed and the spectral resolution combination of such an imaging spectrometer surpasses all other imaging spectrometers I have read about but the low spatial resolution of 8x8 is of course a drawback of the instrument The short maximum recording time is also a limitation of the instrument Villemaire presentation Villemaire The unit most widely used for frequency in spectrometry is the a kayser k The word kayser is not widely used usually substituted with reciprocal centimeters 1 cm equals 2 9979 10 Hz 30 GHz Beer 16 Spatial Fourier transform spectrometry For the Eglin imaging spectrometer above the interferograms are sampled in time this way of acquiring interferograms is called temporal FTS There is another fundamentally different way of acquiring the interferograms
20. 0 000 40 000 USD email Coulson This cost would be way too high but the existing dewars could probably be modified to work vertically The construction that was finally chosen Alternative 3 is depicted in Figure 10 It is capable of collecting spectral images but the ranges of O and are not adequate for the target following feature explained in chapter 4 1 The construction can be easily modified to Alternative 1 or 2 When the chosen construction has been tested it could be further built to work vertically If the Alternative 1 construction would be prefered later the rotating mirror must be replaced with a fixed 45 mirror and the rotating mirror must be placed below the 45 mirror 6 4 50 up PEAREN AEEA EAEN tele e spectrometer o 90 90 Figure 10 Main construction Alternative 3 The spectrometer and the support for the mirror turning system are standing on two beams This construction can be modified to main construction Alternative I or 2 To sum up if the chosen construction would be modified in the future to the standing arrangement Alternative 2 no extra work has been done compared to choosing Alternative 2 from the beginning If the chosen construction would be modified to Alternative 1 in the future there is a little bit of extra work compared to choosing Alternative 1 from the start But if the target following feature will never be used chosing Alternative
21. 1 from the start would have been a waste of time and money Also it is doubtful if the construction according to Alternative 3 could be realized with the available resources of the ScanSpec project see chapter 4 2 5 3 Definition of angles The definition of the following angles will make it easier to explain the scan pattern of the mirror treated in chapter 5 4 Q and 9 are defined in Figure 10 02 and 0 are defined as 2 0 82 90 When 2 and 9z are close to zero they approximate the azimuthal and the elevation angle for the chosen construction 2 is zero when the line of sight and the optical axis of the telescope are orthogonal 2 is posistive if these angles are acute 2 is the elevation angle when 2 0 The rotation axis of the mirror turning system that changes the 2 angle will be called the 2 rotation axis The other rotation axis will similarly be called the 0 gt rotation axis 22 x and y are defined as x 62 2 and y 02 These angles are the mechanical angles that the mirror turning system must turn the mirror to accomplish a line of sight with the angles z and 02 The 2 in the expression above is because the mirror only has to turn the angle A 2 around the z axis to change with the angle Ado 5 4 The scanning of the mirror The specifications of the mirror turning system depend on how the mirror is scanned to acquire a spectral image There are basically two different ways of doing this One way i
22. 2000 tan 0 28 mm 260 mm Since the angle of incidence i is not zero the mirror has to be larger than 260 mm Figure 1 shows the situation The angle of incidence i will vary between 20 and 47 which gives a range of the 2 angle of between 50 and 4 towards scene towards telescope intersection of the rotation axes mirror surface gt a 260 mm Figure I 62 Appendix 2 Geometry gives that d tani yli 2 2cosi and x i dtani 2 cosi We know the distance a The distance d depends on the rotary table and how the mirror is mounted We assume that d will be between 80 and 100 mm which it is in the realized construction The maximum values of y i and x i when i varies between 20 and 47 and d varies between 80 and 100 mm are Ymax 110 MM Xmax 298 mm 15 mm have been added to these values so there will be a 15 mm border around the edges of the used area to compensate for misaligment The other dimension of the mirror only needs to be 260 2 15 mm 290 mm The dimensions of the mirror are shown in Figure 2 290 125 313 438 Figure 2 The dimensions of the mirror 63 Appendix 3 Appendix 3 The standard scan programs This is a list of the parameters of the eleven standard scan programs The first two parameters are spectral resolution and image size Standard scan program 1 NORMAL 8 16cm 8x8 4x spatial over
23. 8 2 1 What isa SPEC Al ama ce soiosvsskeskesivserskersssiseetsbeterelatersseisseie ieee et eee aE 8 2 2 The history of spectral 1MAages eee eeeesseeeeeesseeeceesnaeeccesseeecessaeeeessaeessessaeeesessaeeeeees 9 2 3 The military use of infrared spectral images ssmssssssssrrrrssrrrrrssrrrrssrrrrssrrrrssrrrrrrsrrrrs sens rss r rr n na 10 2 4 Spectral resolution and frequency range needed for target detection ssssessssrsrsssne 10 2 53 Conclusion PEE AE E A 11 3 Imaging spectrometeis ssns soises cessie o eosi os ei oiee eisa ois sios 12 Del AMMO CUCHION 5 5 555 555 Satsihs ses haste T shapeedsteephibeMaghitahi ashes MoeDhedhens odhco dd edhe ed er Gere 12 3 2 Imaging spectrometers based on filteriNng ssesssrrssssrrerrrssrrrrssrrrrssrrrrssrrrrrsrrrrrsrrrrr ss rr rr r rna 13 3 3 Imaging spectrometers based On dispersion s ssssssrrssssrerrsssrrrrssrrrrssrrrrssrrrrrsrrrrrrnrr rss rr enn r rna 13 3 4 Imaging spectrometers based on Fourier transform methods sssssssreresssrrrsssrrrrssrrrrsnrrrn 15 PART II BUILDING SCANSPEC 4 The goals of the project oosssssssoooossssssssoooo0ssseoss000000000000000000000000000000000000000000000000 0000 18 AN The SOAS nionee eheee nee ee e 18 4 2 THE TESOUTCES er EE E occ EEEE EAE Siem co ev ERE den EAEE EERENS 18 5 The overall design of ScanSpec ccsssssscsssssssccsssssscsssssccscsscsscscsscsscsssscsssessssssscsssscesers 19 5 1 The Bomem MR254 FT
24. Eismann and C R Schwartz Focal Plane Array Nonlinearity and Nonuniformity Impacts to Target Detection With Thermal Infrared Imaging Spectrometers Proceedings of the SPIE Vol 3063 pp 164 173 1997 Eismann3 M T Eismann et al Infrared Multispectral Target Background Field Measurements Signal and Data Processing of Small Targets Proceedings of the SPIE vol 2235 Orlando FL USA 2235 09 April 1994 Eismann4 M T Eismann et al Target Detection in Desert Backgrounds Infrared Hyperspectral Measurements and Analysis Signal and Data Processing of Small Targets Proceedings of the SPIE vol 2561 San Diego CA USA July 1995 Gower J F R Gower G A Borstad C D Anger H R Edel CCD based imaging spectroscopy for remote sensing the FLI and the CASI programs Canadian journal of remote sensing Volume 18 Number 4 October 1992 Hackwell John A Hackwell et al gt LWIR MWIR Imaging Hyperspectral Sensors for Airborne and Ground Based Remote Sensing SPIE vol 2819 August 1996 Hammer Philip D Hammer Francisco P J Valero David L Peterson William Hayden Smith Remote Sensing of Earth s Atmosphere and Surface Using a Digital Array Scanned 49 Interferometer A New Type of Imaging Spectrometer Journal of Imaging Science and Technology vol 36 number 5 September October 1992 Hinnrichs Michele Hinnrichs and Mark Massie New approach to imaging spectroscopy using diffractive
25. FOA R 99 01041 615 SE February 1999 ISSN 1104 9154 Technical Report Frans Lundberg The design of ScanSpec a scanning imaging FTIR spectrometer Division of Sensor Technology SE 581 11 LINKOPING DEFENCE RESEARCH ESTABLISHMENT FOA R 99 01041 615SE Division of Sensor Technology February 1999 P O Box 1165 ISSN 1104 9154 SE 581 11 LINKOPING SWEDEN Frans Lundberg The design of ScanSpec a scanning imaging FTIR spectrometer Distribution FMV FlygES FMV FlygFP FOA 33 74 Issuing organization Document ref No ISRN Defence Research Establishment FOA R 99 01041 615 SE Division of Sensor Technology Date of issue Project No P O Box 1165 February 1999 E3016 SE 581 11 LINKOPING SWEDEN Project name abbrev if necessary Optical Signatures Author s Initiator or sponsoring organization FM Frans Lundberg Project manager Frans Lundberg hotmail com fars Bokmal Scientifically and technically responsible Claes Nelsson Document title The design of ScanSpec a scanning imaging FTIR spectrometer Abstract This report describes the ScanSpec project at the Department of IR Systems at the Defense Research Establishment FOA in Link ping Sweden The purpose of the project was to build an imaging spectrometer It has been done by connecting a narrow angle telescope to a fast FTIR spectrometer and by rotating a mirror in front of the telescope This way spectra of IR radiation from different d
26. Figure 18 The placement of the three mirror mounting points are also shown The complete AutoCAD drawings are available on the CD CD ScanSpec98 1 6 4 The support construction The support construction that keeps the mirror turning system and the spectrometer in place is based on aluminum U beams with the dimensions 7 x 140 x 60 mm The whole contruction can be seen in Figure 19 33 Figure 19 AutoCAD drawing of the complete construction The distance between the beams is large enough to make it easy to modify the current construction to the main construction Alternative 1 as depicted in Figure 8 page 21 6 5 Communication between ScanSpec parts The communication between the different parts of the ScanSpec system will be desribed in this section The information paths of the ScanSpec system are shown in Figure 20 The physical information paths the wires between the MM the motion controller MM4005 and the rotary tables RV120PP and RV160PP and between the MM and the joystick were already supplied The only paths implemented by me are marked with thick arrows in the figure The wiring for the hardware switches is very simple It is only a circuit that is closed as long as none of the hardware switches is activated If one of the switches is activated the motor power for both motors is immediately turned off RV120PP RV160PP Hardware switches Figure 20 The flow of information in the ScanSpec system
27. IR spectrometer ssessssresssrrerrssserrsssrerrrssrsrrrssrrrrssrrrrssrrrrsnsrrrrrnrrr rn na 19 32 Main S 02a NEEE A ATAT TAE TAT TATATA 20 5 3 Definition OF aN SlES Sosro 00000 NANNA 22 5 4 The scanning ofthe Miotoine aee eee ao ee roa eere oeta tee tet 23 6 Th ScanSpec hardwaren ea os eS e E E E e ee aeaee 26 OTe W 212 0001 10 CAR quem am E AA R E A EE ERORE REREN 26 622 The miror MOUNNN Ec iesse iritira ioi i EAE RE TE a EET EEA EREE EES Dees 28 6 3 The mirfof turning SYStell 0sessscessssbscesssesisesisesisssidesiesdssniseassretidarreresisesisesises ids irAr isen ENN 30 6 4 The support construction s seesseressssrrerrssrrrressrrrrssrsrrrssrrrrsssrrrrssrrr rss rer s rss rss ser rss ser ss r rr rn ren 33 6 5 Communication between ScanSpec PaltS sseesssersssssrrrrrrrssssrrrrrrrrrrssrrrrrrrrrrrsnrrr rr rr rr nro nr 34 7 The SCanSpec SOflWALCisssorsvosevsrosvosssssossrsvorsrsrsssssesessssssssrosessrssessssssessrosersrssssnssssssessssnsseserere 36 P E EEE e T esc eesti so a a SS OS A 36 T2 Signal PrOCeSSm Ge sss ssccssecegecevecascobscsdueesusdseesevetsecssvecdecesvecdecceuscdacceusdcecseestoesedesuceadiectsndeds 37 7 3 Testing reliability and the timing problem sssseressssrsrrssrrrrrssrrrrssrrrrssrrrrssrrrrsnsrrrrrnrrr rn na 40 PART III WHAT SCANSPEC CAN SEE 8 The first measurements ssssoosoosssssssoooosssocccsersooo0sso0s00000000000000000000000000000000000000000000 0000 0000 8 1 Measurements with th
28. a al Figure 24 The measured and interpolated spectral radiance values Figure 24 part B shows an enlarged section of part A The black values are the measured values The blue values are the interpolated values and the red values are the final pixel values The blue function is filtered with a low pass filter with a cut off frequency of 5 cycles per degree The filter 39 was implemented with the Matlab Signal Toolbax function fir1 The filter is a finite impulse response filter made with the window method The window is a Hamming window The order of the filter is two times the spatial oversampling This way of reducing the noise is better than just averaging four radiance values in the sense that the motion blur is reduced It can be discussed though if the motion blur is so serious that this rather complicated signal processing is necessary Initial testing shows that the motion blur if no spatial oversampling is used is not as serious as anticipated 7 3 Testing reliability and the timing problem Testing Since there are so many different combinations of input parameters for the ScanSpec software system there is no way of testing for bugs for all possible cases of input to the programs Instead the eleven standard scan programs have been chosen and tested All the standard programs work but the timing problems discussed below are worse for certain scan programs Reliability There are four things that affect the relia
29. al axis of the telescope The focusing range of the telescope and the field of view of the telescope detectors have also been measured Only the data from the measurements are presented without conclusions How the measurements were done The narrow angle telescope Bomem SMY02 were mounted on the spectrometer A blackbody source was set at 1200 C at a distance ranging from about 40 to 90 m from the telescope A grid was drawn on a piece of cardboard with a hole The grid was put up in front of the blackbody source and the hole was aligned with the aperture The spectrometer was standing on a tripod that could be panned and tilted The CCD camera of the spectrometer was connected to a monitor A mark was made on the screen of the monitor The spectrometer could then be turned up down left and right The grid in front of the blackbody source together with the screen could be used to measure the angles the spectrometer was panned and tilted The mark on the screen was made in the center of the visual field of view through the CCD camera when Aperture AB see next paragraph was set at 1 1 mm The spectrometer has three apertures one directly after the telescope will be denoted Aperture AB and one for each of the two detectors The aperture for the MCT detector will be denoted Aperture A and the one for the InSb detector will be called Aperture B There are several diameters of each aperture 6 4 4 5 3 2 2 2 1 6 1 1 and 0 8 mm The tempe
30. ariables should tell where the spectral radiance is measured and the spectral variable should tell for what frequency wavelength it is measured The spatial variables can be distances or angles The spectral variable is either frequency or wavelength For ScanSpec the spatial variables are angles and the spectral variable is frequency The spectral radiance Lspec is often defined as dL a ees Equation 1 meek dh where L is the radiance and A is the wavelength Physics p 264 Pedrotti p 265 But when the frequency fis used as the spectral variable the following definition can be used dL Equation 2 spec f df q There are several reasons why I have chosen to use frequency as the spectral variable when acquiring the spectral images The most important reason is that the spectrometer in the ScanSpec system Bomem MR254 records spectral radiance according to equation 2 It is also often easier to work with frequency instead of wavelength since the energy difference between two photons with a fixed frequency difference is always the same Also unlike the wavelength the frequency of the electromagnetic radiation is independent of the medium the radiation is propagating in A spectral image can be seen as an image where each pixel is a radiance spectrum instead of just a gray scale value as it is in an ordinary black and white image Throughout this report the word light will be used for both the visible and the inf
31. as possible but not longer than 8 10 weeks The first approach was to find a complete system to buy which would be as complete as possible only leaving me with the mirror mounting to do But finally separate parts were bought since there was no complete mirror turning system that met the requirements above and at the same time was within the budget of the ScanSpec project Many alternatives were considered more or less in detail Some of them are briefly described below Generally the companies are well represented on the internet and references to the internet addresses are given Alternative Sagebrush Sagebrush internet Sagebrush offers a gimbal system Model 20 Pan amp Tilt Gimbal that is inexpensive around 80 000 SEK but unfortunately the torque is too low for the ScanSpec application Their gimbal with a camera mounted is shown in Figure 14 Sagebrush is working on a new larger gimbal but it was not available for sale in June 1998 Figure 14 The gimbal system from Sagebrush 30 Sagebrush could also design a custom made gimbal for the performance needed but they said they could not possibly deliver such a system within the desired delivery time Alternative Daedal Contacts were made with Ostergrens elmotor a Swedish representative of Daedal Inc regarding the making of a custom designed gimbal system for our needs They tried to design a system but failed to do so before the very last possible date for an order fr
32. at least 20 C would be desirable when calibrating the system A black body that keeps the same temperature as the surrounding environment could also be adequate when measuring cold targets 43 Figure No 1 File Edit Window Help sim file info more ScanSpec data collimator test object prog6 si D im 7 7147cm 1 no spatial oversampling phi2_LL 0 2 pease 8 FOV_pixel 0 1 resolution chosen 16em 1 image size 32 2 ANBI spectrum any CEES EE TEE E a Figure 25 The collimator test object acquired with standard scan program 6 Figure No 1 File Edit Window Help simfile info more D ScanSpec data MRTD measurements deltaT 0 2 long sar ime sim imple tir 7 7147cm 1 2 times spatial oversampling phi2_LL 1 4 theta2 te 0 5 FOV_pixel 0 1 resolution chosen 16cem 1 image size B 2 _ ve spectum appy ate spectum amne BEN Ce Boe ff of Figure 26 Minimum resolvable temperture difference MRTD measurement The temperature difference between the target and the background was 0 2 C 256 interferograms were averaged per measured spectrum Two times spatial oversampling was used The image to the right shows a similar measurement when the temperature difference was 20 C 44 Figure No 1 Figure 27 The Volvo 440 seen in two different frequency bands The car was running and the fan was on In the right window the headlights are the most dominant feature Figure No
33. atures of backgrounds and potential targets The measurements can be used to build models of the IR radiation from vehicles and other objects of military interest Apart from research spectral images can be used for surveillance and targeting Radar sensors provide a wide area coverage but for a given detection probability they currently have unacceptably high false alarm rates Eismann Therefore the use of visible and infrared sensors must be considered to decrease false alarm rates Other disadvantages with radar compared to passive optical sensors are that a radar transmitter is easier to detect than a passive optical detector and a radar can be jammed During the day the visible and the near infrared parts of the spectrum are interesting but when it gets dark MWIR and LWIR sensors become more important Broad band infrared imaging sensors for MWIR or LWIR IR cameras have been used successfully to detect hot targets in cold backgrounds For example a person in a cold background is easy to detect Camouflaged targets in warm backgrounds are harder to detect with broadband IR sensors For such targets spectral as well as spatial information about the radiation is needed in order to distinguish the target from the background Eismann Eismann and Schwartz 2 4 Spectral resolution and frequency range needed for target detection Eismann Schwartz Cederquist Hackwell and Huppi have investigated which wavelength bands that are most
34. bility of the system 1 Bugs in the software written by me The software has been tested with the standard programs with no bugs found 2 The reliability of the software Acquire Acquire has a tendency to crash when large amounts of data are collected especially when over about 32 Mbyte of data is stored at the Bomem acquisistion board 3 Problems with data transfer to the MM On some occations the data transfer from the FTIR computer over the serial cable to the MM has been malfunctioning When the standard scan programs were tested with the final software version 98 1 there was no problem with the data transfer 4 The timing problem This problem is described in the next section The timing problem The MM is synchronized with the FTIR computer by a trigger signal sent from the MM to the parallel port of the FTIR computer The parallel port is checked with a C program written by me There are of course certain delays when sending the trigger and reading it However intitial testing shows that the timing error can be as large as 0 05 s varying from time to time and this is probably not only becasue of these delays Since many programs must be run on the FTTR computer at the same time at least Acquire Matlab the C program to check the parallel port and the operating system Windows 95 I believe the probable cause of the timing problem is that the Pentium processor of the FTIR computer is not necessarily running the C program a
35. canSpec project That would also have meant more code and code harder to maintain More code is not desirable since the code already consists of 4300 lines 127 000 characters The hardware and the software have been constructed with consideration to possible future changes 46 The costs of the project except for my salary are Two rotary tables with motion controller 160 000 SEK The mirror 2 000 SEK The support beams 2 000 SEK Workshop cost 30 000 SEK Other expenses lt 3 000 SEK Which sums up to around 200 000 SEK which was the approximate budget of this project so the economic goal was met The goal of finishing the project within 20 weeks of work for one person and within a total time of 24 weeks was not accomplished It took 23 weeks of work within 28 weeks The reader can judge if the project has been well documented in this report and in CD ScanSpec98 1 9 3 Possible improvements changes and new applications ScanSpec is ready to be used in field measurements but important improvements could be made Some of them are mentioned below The possible improvements of the software have already been discussed The system could be more prepared for transportation and field use to make it more practical to make measurements The system can be modified for totally new applications For example the noise levels for different frequencies and signal levels could be measured automatically with ScanSpec This can be done by movin
36. ce is placed on a separate line it is for the whole text between the last headline and that reference The name within brackets is a keyword used to find the reference in this section If the reference is not a reference to printed material there is also a word before the keyword describing the type of the source Articles books and other printed material Beer Reinhard Beer Remote Sensing by Fourier Transform Spectrometry John Wiley amp Sons Inc 1992 Physics Carl Nordling and Jonny Osterman Physics Handbook 4th edition Lund Sweden Studentlitteratur 1987 Bomeml FT IR spectroradiometer catalog Bomem Inc April 1994 Bomem2 MR series FT spectroradiometers design overview and theory Part of the documentation of the Bomem MR254 spectrometer Bomem Inc May 1995 Breckinridge James B Breckinrigde Evolution of imaging spectrometry past present and future Proceeding of the SPIE Vol 2819 pp 2 6 August 1996 Eismann M T Eismann C R Schwartz J N Cederquist J A Hackwell R J Huppi Comparison of infrared imaging hyperspectral sensors for military target detection applications Proceedings of the SPIE Vol 2819 pp 91 101 August 1996 Eismann and Schwartz M T Eismann and C R Schwartz Focal plane array nonlinearity and nonuniformity impacts to target detection with thermal infrared imaging spectrometers Proceedings of the SPIE Vol 3063 pp 164 173 1997 Eismann2 M T
37. celeration without bending too much Requirement 1 can be obtained by mounting the mirror at only three points Theoretically no moments can then act on the mirror due to the mounting In reality there are no ideal points but the the points should be made as small as possible Also an elastic material rubber can be used to minimize moments acting on the mirror through the mounting points The construction of one of these three mounting points can be seen in Figure 13 I chose to use adhesives to fasten the mirror Another way would have been to use clips around the edges The adhesive method has the advantage that the mirror can be fastened anywhere on the backside of the mirror not only at the edges The drawback is that once glued nothing can be changed 28 glue surface mirror TN brass piece threaded for i the screw aluminum plate 8mm rubber washers screw Figure 13 A drawing of one of the three mirror mounting points Requirement 2 has been proved to be fulfilled with this construction and the scan pattern used Requirement 3 The three mounting points must be located to minimize the bending of the mirror caused by forces This is a problem that has not been treated in detail An analythic calculation was made on the one dimensional case bending of a beam but it might not be valuable for this case The problem is well defined though a
38. corded by an imaging spectrometer can be viewed as a data cube of numbers of the spectral radiance in three dimensions two spatial dimensions x and y and one spectral dimension f A 2D array of IR sensors can obviously record only two of these three dimensions at the same time The third dimension the spectral or one of the spatial ones must therefore be sampled temporally If a 1D array of IR sensors would be used two of the three dimensions would have to be sampled temporally If only one sensor is used as is the case with ScanSpec all the three dimensions of the data cube must be temporally sampled Since the performance of 2D arrays of IR sensors have recently been improved much and is being improved rapidly I have no doubt that currently the best way of constructing imaging spectrometers is with a construction based on a 2D array of I R sensors Therefore the following text will only treat imaging spectrometers based on such sensors There are basically three different ways of extracting spectral information from infrared radiation 1 filtering 2 using dispersive instruments 3 using Fourier transform instruments These three ways can all be used when designing an imaging spectrometer and will be exemplified I will not try to give a complete summary of what different types of imaging spectrometers there are and their advantages and disadvantages The following should be seen as examples of the instruments that exist today and the funda
39. ctral resolution of AMA 0 01 is used with hyper spectral sensors At this resolution the radiation from solid and liquid surfaces reveals details of the chemical composistion Ultra spectral instruments that can record spectral information in a large 1000 to 10000 number of spectral bands are called ultra spectral The spectrometer used with ScanSpec is ultra spectral if the highest spectral resolution is used With such a high spectral resolution the chemical composition of gases can be studied According to Breckinbridge s computations Breckinbridge a resolution of 10 nm approches the optimum when studying solid objects in the range 0 4 to 2 5 um For comparison the Bomem MR254 FTIR spectrometer used in the ScanSpec system has a maximum spectral resolution of 0 1 nm at I um and 10 nm at 10 um Breckinridge also explains that lower resolution is needed for solids than for gases because the atoms and their electronic transitions which form the lines are more tightly bound in the solid lattice and thus give much broader spectral features than those spectral features produced by gases Hackwell states that a resolving power A AA of 200 resolves any spectral features seen in a solid or liquid surface 2 5 Conclusion IR cameras can detect hot targets in a cold thermal background but in order to detect targets with a temperature not much different from the background imaging spectrometers are necessary The best wavelengt
40. detector which is used to detect longer wavelengths The region for which the spectrometer is sensitive is nearly circular but the sensitivity within that circle varies from 1 0 to 0 3 arbitrary units The sensitivity within the field of view of the InSb detector that is shown in figure 3 is much more uniform The sensitivity within the central part of the circular field of view varies between 1 0 and 0 75 arbitrary units 54 Appendix 1 0 1 0 05 0 0 05 0 1 0 15 Ao deg Figure 3 The normalized sensitivity of the spectrometer within the field of view The narrow angle telescope Bomem SMY02 and the InSb detector were used Figure 4 and 6 show the magnitude of the Fourier transforms of the angular sensitivity functions in Figure 2 and 3 Figure 5 is the scale for those figures 55 Appendix 1 fe 7 cycles deg f cycles deg Figure 4 The magnitude of the two dimensional Fourier transform of the angular sensitivity the MTF for the MCT detector The values are normalized so that the maximum value equals 1 Figure 5 Scale for figure 4 and 6 56 Appendix 1 fo cycles deg 15 10 5 0 5 10 15 fy cycles deg Figure 6 The magnitude of the two dimensional Fourier transform of the angular sensitivity the MTF for the InSb detector The values are normalized so that the maximum value equals 1 Errors The statistical errors in the measurements were estimated by measuring the sens
41. e InSb detector and above 82 for the MCT detector so aliasing should not be a problem if the sampling frequency is 10 samples degree The scan pattern first implemented was the one in Figure 11 The effective scan lines goes both in the right and the left direction 0 Figure 11 Scan pattern The effective scan movements are marked with thick line The background grid shows the locations of the pixels in the spectral image That type of scan pattern showed to work poorly because of timing problems Instead the scan pattern shown in Figure 12 was used with better results Note that when the dotted parts of the path in Figure 12 is traced the mirror moves as quickly as possible For most scan programs the ineffective scan time is small compared to the effective scan time The ineffective scan time is the time when the line of sight traces the dotted part of the path and the effective scan time is the time when the line of sight is tracing the part of the path marked with thick arrows 1 i The user can choose a scan program which consists of parameters that control the scanning of the mirror such as image size and spectral resolution 24 Figure 12 The scan pattern used The mirror is scanning towards the right only The background grid shows the locations of the pixels in the spectral image Thick arrows indicate the paths when the spectrometer is acquiring useful information The scan pattern in Figure 12 i
42. e standard scan programs sssssssssrrrrsssrrrrssrrrrrerrrrssrrrrrsrrrrsrsrrrrrnnr rna 82 MRED Me aSUreMents sess os scceak cc hescdeds casededse esh oeobas ou bec eheas fade guadeegwob ole veduesddeesoncdivasesedgeaet nees 8 3 Measurements of a VOLVO AA uusiuisssedidssrsedeskd derk dera es Mea dere riderr OA Cali DratiOM v E EEEE Etta cdc sas ets E Nido ta dy 9 Results and possible improvements csssssssssscsssssssssssssccccssssssssssscsssscssssssssssssssssssess OT gt IDITST TTI T s sssecasseea testcase eae ea dea aa NT AOA 10 CONCIUSIONS oicseeccessesectesctessteuctes steed sasscessdeusteseveacdeusseustsesdeacteussevatsuedeaceenscevatsesdescdeussevatsesteaces MUS EE TE AA senere sosse cosecoseaone sets cone ccnunetaest ences nanauan esate sone saneaneeene sone sasevanewenesene neues Appendix 1 Measurements of field of view angular sensitivity and focusing range Appendix 2 Calculation of the mirror size Appendix 3 The standard scan programs 1 Introduction Imaging spectrometry has developed rapidly the last years and has applications in many fields of science The military applications include measuring optical signatures surveillance and targeting The ScanSpec imaging spectrometer described in this text is designed to be able to measure optical signatures of backgrounds and potential military targets This report will tell the reader about the ScanSpec project which was done at the Department of IR syst
43. e sufficient for this target following feature ScanSpec should be documented so that it can be used with no further information than the User s manual There should also be a report describing the project this report Both the hardware and the software should be constructed to make it easy to modify for other new applications 4 2 The resources The ScanSpec was to be built tested and documented using 20 weeks of work for one person and within a total time of 24 weeks The cost of the project could not exceed approximately 200 000 SEK not including my salary 18 5 The overall design of ScanSpec 5 1 The Bomem MR254 FTIR spectrometer To understand the design of ScanSpec a few things need to be known about the Bomen MR254 spectrometer Only the information needed to understand the rest of this report is provided here For further information please refer to Bomem1 and Bomem 2 The Bomem MR254 is an FTIR spectrometer Fourier transform infrared spectrometer with changable spectral resolution The spectral resolution of 1 2 4 8 or 16 cm can be chosen The standard programs of ScanSpec use a resolution of 4 or 16 cm The spectral range of the spectrometer itself is from 0 7 to 20 um but the usable range depends on what detectors that are used The Bomem spectrometer used for ScanSpec is equipped with two IR detectors an InSb and an MCT detector The optical layout of the spectrometer makes it possible to use both the
44. ed and constructed in Canada in 1981 It uses a CCD camera covering the visible and near infrared part of the spectrum The first entrance lens for the FLI forms an image on a slit The second lens collimates the light Then the light passes through a transmission grating and the last lens before it is detected by a 2D array of light sensors Gower There are imaging spectrometers for the infrared region that are based on the same basic principles as the FLI above for example SEBASS SEBASS works in the MWIR and the LWIR regions More information about SEBASS can be found in Hackwell and Eismann 13 An imaging spectrometer that disperses the light along the optical axis Pacific Advanced Technology Inc California USA has used a new approach to dispersive imaging spectrometry that has been patented A single optical element is used see Figure 3 That element performs both the imaging and the dispersion An 2D array of light sensors is moved as indicated in the figure Only one specific wavelength of light is focused on the sensor array at a time The optics is designed so that the depth of focus is very shallow allowing for high spectral resolution the imaging and dispersive optical element sensor array s incoming light gt Figure 3 The principle of the Pacific Advanced Technology imaging spectrometer Light of longer wavelength will be refracted more in the optical element than that of s
45. ems at the Defense Research Establishment FOA in Link ping Sweden The goal of the project was to build an imaging spectrometer based on a fast non imaging spectrometer This has been done by connecting a narrow angle telescope to the spectrometer and by turning a mirror in front of the telescope In this way the infrared radiation from different directions can be collected By scanning the mirror around two rotation axes a whole spectral image can be acquired A spectral image is an image in which each pixel consists of a spectrum The imaging spectrometer is named ScanSpec and the finished construction can be seen in Figure 1 Figure 1 ScanSpec The system is based on a fast FTIR spectrometer Fourier transform infrared spectrometer from Bomem Inc The spectrometer is capable of collecting 65 spectra s at the highest speed and lowest spectral resolution The report is divided into three parts plus this introduction and the conclusions chapter Part I Spectral images explains what a spectral image is its history and how spectral images can be acquired Part II Building ScanSpec will tell the reader how ScanSpec was built and the last part part III What ScanSpec can see will present what ScanSpec can be used for and its performance That part also presents the initial measurements including the world s probably first hyper spectral image of a Volvo 440 This project is the last part of my masters degree at the University of Umea
46. ered A thorough investigation of how the MRTD depends on these factors has not been made The minimum resolvable temperature difference is the minimum temperature difference between a test target and a background both black bodies for which the target is percieved Holst page 7 42 8 3 Measurements of a Volvo 440 The first real object measured with ScanSpec was a Volvo 440 The distance between the car and ScanSpec was 70 m One spectral image of the car is shown in Figure 27 As an example of how different features of the car are predominant in different frequency bands a band where almost only the headlights can be seen was chosen The car including the exhaust is seen in Figure 28 The exhaust emits a lot of IR radiation around the frequency 4100 cm The original 32x32 image has been interpolated using splines to create a 125x125 image If the calibration of the spectrometer is done carefully someone with knowledge about spectrometry could get much of information from images like this For example the gas contents of the exhaust could be studied and probably some conclusions about the features of the headlights could be drawn from the data 8 4 Calibration The calibration of the ScanSpec system is a bit tricky since the entrance optic is large There is a large area blackbody source at the Department of IR systems at FOA that can be heated but not cooled To study cold winter backgrounds a blackbody that can be cooled to
47. esolution image size time to post process the data and time to acquire the spectral image The ScanSpec software and maybe the MM limits how low the image distortion could be The ScanSpec hardware limits the range of the sky that can be measured but it is built to be easily changed to cover a larger part of the sky The reliability of the system is limited by both the spectrometer Acquire often crashes for large amounts of data and the other parts of the system the timing problem and problems with data transfer to the MM see chapter 7 3 page 40 The minimum performance goal of being able to acquire a 10x10 spectral image within a minute at the lowest resolution of the Bomem spectrometer is met and far surpassed ScanSpec can acquire an image with 10 times more pixels 32x32 in around 30 s Any software system can be improved The software for ScanSpec has been written to be easy to use and change Possibility of re using the code has not been considered since the application is very specialized with only very few parts that could be used in other software systems The error checking is not at all complete but is still a significant part of the code The software limits the total performance of the system to a number of fixed image sizes and a pixel field of view of 0 10 It can only look one way not both to the right and the left of the spectrometer The software could have been made more general but not within the time goal of the S
48. f the times between the spectra In Figure 24 the spatial oversampling is 4 and the size of the spectral image is 8x8 pixels Part A of the figure shows the spectral radiances recorded The radiance at the positions on the 2 axis that are marked with blue lines will be estimated by linear interpolation and extrapolation Linear interpolation and extrapolation was chosen since this signal processing has to be quite fast Note that the procedure described here for only one scan line and frequency must be repeated for all frequencies and scan lines Now we have radiance values at all the positions marked with blue lines in Figure 24 but we only need values at the positions marked with red lines which are the centers of the spectral image pixels We know that the telescope works as a low pass filter on the spatial frequencies of the studied scene Very roughly the telescope destroys all the spatial frequencies above 5 cycles per degree see Appendix 1 therefore we know that all the frequencies in the measured radiance signal above 5 cycles per degree is due to noise Let us remove those frequencies with a filter to reduce the noise spectral radiance E EN SN 7 A sea A Bie T T PE A AE TALAT HEA red bam SSS mi r e 7 red blue B M easured values black black Interpolated values blue Pixel values red blue 7 red __ L
49. f the 400 times are not what they should be The resolution 8 cm was used two interferograms were added per measurement oversampling was on and both detectors were used Unfortunately the times when the spectra are acquired cannot be controlled at all When the acquisition has started the spectrometer cannot be controlled by any external events Another timing problem is that if the spectrometer is for example set at the resoluton 8 cm and set to add two interferograms per measurement it should be able to acquire 25 spectra s Usually it does but not always In most cases the time between two consecutive spectra is 1 25 s but sometimes that time is 1 5 25 2 25 or even 3 25 s Figure 7 shows what it can look like This irregularity and the fact that the spectrometer is independent of external events once it has started has made the signal processing much more complex than what it could have been 5 2 Main design To build an equipment that complies with the target following feature explained in chapter 4 1 the first type of construction considered was the one in Figure 8 20 mirror A up i telescope spectrometer 90 90 lt __ gt mirror B Figure 8 Main construction Alternative 1 The mirror A is fixed at a 45 angle with the optical axis of the telescope The mirror B can be rotated 90 around the vertical axis as indicated and it can also be rotated around a horizontal axis th
50. g is not a new technique but it should not be underestimated compared to the various new approaches especially as new filters develop Conceptually the design is simple the instrument can be built small and light weight and is not too expensive 3 3 Imaging spectrometers based on dispersion As we will see there are several completely different ways of designing a spectral imager based on dispersion Gratings or prisms can be used to spread out the wavelengths in space I will not try to give a complete description of all different approaches that exist I will merely give a few interesting examples The examples will include commercially available imaging spectrometers Simple imaging spectrometers based on dispersion The simplest type of imaging spectrometer based on a 2D array of light detectors uses lenses an entrance slit and a diffractive optical element The diffractive element diffracts the light in one direction The other direction is left for spatial imaging The array of light detectors can simultaneously record the spectra in one direction and one spatial dimension in the other direction of the array The second spatial dimension has to be temporally scanned This can be achieved by turning a mirror in front of the spectrometer by turning the whole spectrometer or by moving it in one direction for example by using it from an airplane One of the first of these instruments to be built is the Fluorescence Line Imager FLD design
51. g the line of sight from a hot to a cold target and for each position several spectra can be acquired By computing the average and the variance of the measured radiance values the signal to noise ratio for different frequencies and signal levels can be estimated The modulation transfer function MTF could also be measured automatically This can be done by measuring the spectral image of a point source But the spatial sampling frequency should then be higher then the two samples per 0 20 that is currently used with ScanSpec The MTF of the system for a certain frequency is the Fourier transform of the radiance image for that frequency Another application would be background subtraction For example when flares from airplanes are measured the flare could be followed with the mirror of ScanSpec using the joystick The positions of the line of sight would be recorded and when the flare has burned out the mirror could trace the line of sight backwards to measure the radiation of the background without the flare The background radiation could then be subtracted from the measurements 47 10 Conclusions An imaging spectrometer ScanSpec has been developed from a non imaging FTIR spectrometer from Bomem Inc The ScanSpec project has been successful in that ScanSpec widely increases the capabilities of the FTIR spectrometer Other imaging spectrometers for the IR region have been developed but there is not much commercially available on the mar
52. gher spectral resolution at a specific signal to noise level than do equivalent aperture grating based instruments Hammer 17 4 The goals of the project 4 1 The goals The goal of the ScanSpec project was to build an imaging spectrometer using an existing non imaging spectrometer Bomem MR254 with a narrow angle telescope This was to be done by using a large mirror in front of the telescope that can be turned around two rotation axes to acquire spectra from different spatial locations The spectrometer used is briefly described in chapter 5 1 The main performance design goal of ScanSpec was that the existing Bomem spectrometer should when possible limit the total performance of ScanSpec By when possible is meant when possible within the given time and money resources of the project see chapter 4 2 below The minimum performance requirement of ScanSpec was that it should be able to acquire a spectral image with the spatial resolution of 10x10 pixels within one minute An extra desirable feature was the target following feature That feature is the possibility to continuously measure the R radiation from a moving object by turning the mirror In order to be able to follow an airplane the line of sight of ScanSpec must turn 20 per second or more Also it is desirable to cover a large part of the sky by turning the mirror The coverage of the sky of an elevation angle of 4 to 50 and an azimuthal angle of 90 would b
53. h bands for a high target background contrast fall in the range 8 5 to 11 um A spectral resolution of 100 nm in that region is enough to detect military targets The Bomem MR254 spectrometer has a higher resolution than necessary for this application It is therefore possible to use it to investigate what the lowest spectral resolution needed is in order to detect military targets in typically Swedish backgrounds This is important to know if a spectral imager with higher speed and higher spatial resolution will be bought or constructed at FOA in the future 11 3 Imaging spectrometers 3 1 Introduction The definition of an imaging spectrometer or spectral imager in this report is simply an instrument able to collect a spectral image There are other definitions for example an instrument capable of simultaneously registering the spectrum components AND the spatial location internet Specim I believe there is no instrument today that can simultaineously register both the spectrum and the two spatial dimensions of the radiation In order to do that a three dimensional array of radiation sensors is needed or a 2D array of 1D arrays or a 1D array of 2D arrays 2D arrays of sensors for R radiation is in a phase of rapid development The advent of large 2D arrays of IR sensors has led to the development of various types of imaging spectrometer designs Some of the possible designs will be discussed in this report The data re
54. he start angles et cetera The signal processing in module 2 can be divided into two separate parts Part I computes the positions of the rotary tables at each time a spectrum was recorded by the spectrometer Part II computes the spectrum for each pixel in the spectral image 37 Part I Computation of the positions of the measured spectra The lower part of the scan pattern of a 8x8 spectral image in shown in Figure 22 The points for which the MM might have recorded positions are marked OOOO O gt 0 Figure 22 The lower part of the scan pattern for a 8x8 spectral image The circles mark the locations of the positions sampled by the MM Only two scan lines out of eight are shown The speed is constant when the thick arrows are traced by the line of sight but not otherwise Let us consider time as a function of position We can then extrapolate linearly using the two samples closest to the edge to find the times at the beginning and the end of each of the eight scan lines each of the eight thick arrows When we have found these times we can throw away the information from the spectra not sampled on a scan line PDA i i i i X i Figure 23 The first part of the scan pattern for a 8x8 spectral image The crosses mark the positions of the tables each time a spectrum is acquired Only two scan lines out of eight are shown The next step is to estimate the positions of the tables when each of the
55. he target the four bars and the background was 10 C The left image in Figure 25 shows the spectral radiance of the test target averaged between 687 and 764 cm and the right image is averaged between 1875 and 1975 cm The radiance is shown using a linear gray scale The unit of the spectral radiance is W cm sr cm in the figure but SI units W m sr Hz can also be displayed To make it possible to study the repeatability four measurements were done on the same target with the same scan program see the directory ScanSpec data Repeatability test sims in CD ScanSpec98 1 The repeatability of the ScanSpec has not been studied much but the repetability for the test case is much higher than anticipated 8 2 MRTD measurements The minimum resolvable temperature difference MRTD has been estimated If many interferograms are added per measured spectrum the MRTD is around 0 2 C Figure 26 shows a test measurement with a target background temperature difference of 0 2 C where I consider the target to be resolved 256 interferograms were added per spectrum and two times spatial oversampling was used when scanning the image Note that the radiance has been averaged in the frequency intervals best in resolving the target If the standard scan programs are used the MRTD is closer to 1 C than 0 2 C The MRTD depends on how many interferograms that are averaged the spatial oversampling and what frequency band is consid
56. horter wavelengths Only light with a specific wavelength will be focused at the sensor array As the sensor array moves as indicated a spectral image can be recorded The collected spectral image needs advanced digital image processing to increase the spatial and the spectral resolution Non linear deconvolution algorithms have been developed for this The advantage of this approach over a conventional dispersive spectrometer is that the whole aperture can be used to collect light instead of only a narrow slit that has a low throughput It also has an advantage over a Fourier transform spectrometer The photon noise shot noise is lower since only light from a small frequency band is measured at a time not the entire bandpass of the instrument as is the case for a Fourier transform spectrometer However the spectral resolution of a Fourier transform spectrometer is generally higher Pacific Advanced Technology offers an MWIR instrument with the following parameters Spectral range 3 0 5 0 um Nominal Spectral resolution AA A 0 0025 0 01 um at 4 um F number f 2 5 at 4 um A turn key system including IR camera and computer would cost around 95000 USD This instrument could be used with an existing IR camera and a computer which would reduce the cost to 45000 USD They also sell a new product covering the range 8 to 10 um which is a bit more expensive email Hinnrichs US Patent 5 479 258 issued Dec 26 1995 14 Figure 4
57. irections are acquired The mirror is used to scan a scene to acquire a spectral image that is an image for which each pixel consists of a whole spectrum The report includes an overview of imaging spectrometers for the IR region and of the military use of spectral images The development of the ScanSpec imaging spectrometer is described and the first measurements are presented The report is written for anyone interested in spectral images and imaging spectrometers and specifically the ScanSpec imaging spectrometer The reader is assumed to have some university level knowledge about physics in general Keywords spectral images imaging spectrometer FTIR ScanSpec Further bibliographic information Language English American ISSN 1104 9154 Pages 64 p Price acc to price list Distributor if not issuing organization Dokumentets utgivare Dokumentbeteckning ISRN Forsvarets forskningsanstalt FOA R 99 01041 615 SE Avdelningen for Sensorteknik Dokumentets datum Uppdragsnummer Box Tl6 gt februari 1999 E3016 581 11 LINKOPING Projektnamn ev f rkortat Optiska signaturer Upphovsman m n Uppdragsgivare FM Frans Lundberg Projektansvarig Frans Lundberg Q amp hotmail com Lars Bohman Fackansvarig Claes Nelsson Konstruktionen av ScanSpec en skannande bildalstrande FTIR spektrometer Dokumentets titel i vers ttning Sammanfattning Denna rapport beskriver ScanSpec projektet vid Institutionen f r IR system p F
58. itivity for the same part of the field of view three times The part measured was AQ ranging from 0 to 0 16 for A9 0 To incorporate most of the errors the telescope was defocused and focused again and the angle A was changed and set back to 0 between each of the three measurements Also some time passed between the measurements in order to incorporate the temperature fluctuations of the blackbody source The background radiation was subtracted from the result and the values were then normalized to a range between 0 and 1 arbitrary units as with the previous measurements Figure 7 and 8 show the results of these measurements 57 Appendix 1 sensitivity og arbitrary units 04L 02L 0 fi 1 fi fi fi 1 in 7 0 002 004 006 008 01 012 014 016 Ao Figure 7 Three measurements of the sensitivity of a part of the field of view for the MCT detector The difference between the three measurements for the MCT detector is at up to 6 of the maximum measured value 0 8 0 7 sensitivity osl arbitrary units 0 5 0 4 0 3 0 2 0 1 0 L L L L L Hs 3 0 0 02 0 04 0 06 0 08 0 1 0 12 0 14 0 16 Ao Figure 8 Three measurements of the sensitivity of a part of the field of view for the InSb detector The difference between the measurements of the sensitivity of the InSb detector is up to 1 4 of the maximum value except at one point for Ad 0 117 the difference is 10 This is due to
59. ize is 1 9 kg The delivery time is less than 2 weeks A 10 mm thick float glass mirror at the IR department at FOA was tested in front of the telescope The CCD camera on the spectrometer showed an image that was not much blurred by the mirror Unfortunately the mirror was not large enough for this application But the result shows that the flatness of a float glass mirror can be sufficient for this application A normal bathroom mirror about 4 mm thick was also tested but gave a very blurred image from the CCD camera Not surprisingly such a mirror is not flat enough for this application Alternative 1 was discarded since the flatness of a float glass mirror is sufficient Also the price for such a mirror is much higher than for the other alternatives The weight would require much from the mirror turning system and the delivery time is unacceptably long Float glass is flat glass produced by solidifying molten glass on the surface of a bath of molten tin according to internet Webster 27 It was decided to order a mirror from Edmunt Scientific according to Alternative 3 The stated flatness is good enough but there was a question of wether the thickness of 6 mm gives enough bending rigidity to keep the mirror flat when forces act on the mirror due to angular acceleration and the mass of the mirror If the mirror according to Alternative 3 would not be sufficient for the application there is no big loss since that mirro
60. ket that covers the MWIR and the LWIR regions The performance of ScanSpec is good compared to other imaging spectrometers in the MWIR and LWIR regions when it comes to spectral resolution and sensitivity but it is slow and the image size is limited The advent of large 2D array of IR detectors has led and will lead to the development of various new types of imaging spectrometer designs For the most important parameters such as acquisition time spectral resolution spatial resolution data post processing time and maximum image size the spectrometer limits the total performance of ScanSpec Improvements of ScanSpec can be made but are probably best to do later after some measurements have been done to see what features that are most important to improve The system has been made to be easy to change The mirror and the mirror turning system can be used for other important applications not only to acquire spectral images ScanSpec is suitable to measure the infrared radiation from backgrounds or anything that does not move Together with IR cameras with higher spatial resolution ScanSpec can be used for measurements to build a data base of I R signatures of backgrounds and potential military targets 48 11 References A note on the referencing system in this report A reference within a sentence applies for that sentence only and a reference directly after the last sentence of a paragraph applies for that whole paragraph If the referen
61. ment 5 Parameters aperture A 3 2 mm 60 Appendix 1 aperture AB 6 4 mm aperture of the blackbody 12 5 mm blackbody temperature 1209 C distance d a 59 1 m Measurements To obtain the corresponding angles a a2 a3 az multiply by 9 70 10 mm MCT detector a 72 mm a 2 70 mm a 3 57 mm a 4 79 mm Result MCT detector with aperture 3 2 mm and distance d a 59 1 m FOV 0 136 Measurement 6 Parameters aperture B 1 6 mm aperture AB 6 4 mm aperture of the blackbody 12 5 mm blackbody temperature 1209 C distance d a 59 1 m Measurements To obtain the corresponding angles a a2 a3 az multiply by 9 70 10 mm InSb detector a i 29 mm a 30 mm a 3 23 mm a s 36 mm Result InSb detector with aperture 1 6 mm and distance d a 59 1 m FOV 0 058 61 Appendix 2 Appendix 2 Calculation of the mirror size The diameter of the entrance optics of the narrow angle telescope for the Bomem MR254 spectrometer is 250 mm The field of view of the telescope is 0 28 according to Bomem Assume that the mirror will be used so that the length of the optical path from the telescope to the mirror is less than 2 m For ScanSpec 98 1 this distance will be much less then 2 m but there might be later modifications in the future that require a longer distance The diameter of the infinite cone from which the telescope collects light at 2 m from the telescope is 250
62. mental principles of their construction The development of imaging spectrometers is very rapid at the moment and the situation in say five years will be completely different More imaging spectrometers will be on the market and they will have better performance and probably cost much less than today One factor that limits the performance of the imaging spectrometers of today is the huge data rate that would be produced by a fast imaging spectrometer with high spatial and spectral resolution For example assume that we wish to produce spectral images with 500x500 pixels and 1000 spectral 12 bands Further assume that we want to record 25 spectral images per second and that each value of the spectral radiance is recorded with 2 bytes This would demand a data rate of 500 pixels spectral band 1000 spectral bands spectral image 25 spectral images s 2 bytes pixel 12 5 Gbyte s I do not think there is any way of taking care of all that data with the current technology at least not for a reasonable price but if the speed of digital data storage units and processors increase as it has been in the past who knows Real time data compression could possibly be used to reduce the enormous data flow 3 2 Imaging spectrometers based on filtering One easy way of constructing an imaging spectrometer is to use band pass optical filters For exampel a filter wheel or a tunable Fabry Perot filter can be placed in front of an IR camera Filterin
63. nd could be solved numerically with finite element methods FEM Requirement 4 is fulfilled The strength of the three mounting points is good in fact many times better than necessary Requirement 5 The mounting should not introduce forces on the mirror due to changes in temperature If the mirror mounting should be able to work between 25 C and 35 C the temperature range AT is 60 K The linear thermal expansion coefficient for aluminum is 23 2 10 K average between 0 and 100 C and for common glass it is 8 10 K Physics The mounting points will not be more than 0 53 m the diagonal of the mirror apart The difference in thermal expansion between aluminum and glass on that distance and the given AT is 23 2 10 K 8 10 K 60 K 0 53 m 0 00047 m 0 5 mm The holes for the screws in the aluminum plate should be made larger then the diameter of the screw to compensate for this The choice of adhesive Two different adhesives were tested Loctite 406 cyano acrylate and a two component epoxy adhesive called Araldit 2012 The Araldit has superior strength with the tested brass glass combination and it was used for the ScanSpec mirror I will not describe in detail the tests that were made but I will state some conclusions The cleaning of the surfaces is very important as is the roughness of the surfaces The brass surface should be made rougher with a file The strength of the glue specified by the producer is very
64. ngles There are eleven standard scan programs for different image sizes spectral resolution spatial oversampling will be explained later and number of spectra added per measurement The module writes a motion control program which is sent to the MM and executed there Together with Acquire this module collects all the data necessary to compute the spectral image The spectral data is saved in files with the Acquire program The positions of the rotary tables are sampled by the MM with a constant time interval These positions are sent from the MM to the FTIR computer when the scanning of the mirror is finished and then they are saved in a file to be used by module 2 Module 2 Computing the sim This module uses the collected data to create the spectral image The spectral image is saved as the file current sim sim stands for spectral image The file format of a spectral image file is defined in the Software manual CD ScanSpec98 1 This module includes signal processing that will be described in some detail below in chapter 7 2 Module 3 Looking at the sim The purpose of this module is to present the spectral image to the user in an efficient way The User s manual filename User s manual txt in CD ScanSpec98 1 describes the functions of this module The module is implemented using a Matlab GUI Graphical User Interface which is a convenient way to produce user interfaces In the chapter The first measurements
65. nufacturers define flatness differently Protective coating A surface coating is needed to protect the surface from scratches and oxidation A bare aluminum or gold surface must not be touched at all Size and shape The mirror size depends on the range of the angles and 0 It also depends on how close the rotating axes will be to the surface of the mirror See Appendix 2 for the calculation of the required mirror size The size of the mirror chosen is 438 by 290 mm rectangular The weights and prices below are based on approximately that size Weight For the ScanSpec application the mirror has to be accelerated quickly it is important to keep the weight low In our case this means the same as not choosing a mirror that is unnecessary thick 26 There are however very expensive mirrors where weight has been reduced by not making them solid The bending rigidity of such a mirror is much higher than that of a solid mirror with the same weight Price The price of mirrors varies a considerably especially with the flatness parameter The price can be reduced a lot if an off the shelf mirror can be bought instead of one specially designed for just this application If the flatness does not have to be very high float glass can be used and then the price of the glass substrate is low since float glass is not polished For a specially designed mirror the two main costs are the price of the glass substrate if float glass is not used and
66. ogram can be acquired There is a one to one correspondence between this interferogram and the spectrum of the light This relationship will not be derived here but it involves a Fourier transform hence the name of this method I refer to Pedrotti for a more detailed discussion of the subject Infrared imaging spectrometers based on Bomem Spectrometers Bomem Inc has been involved in the development of imaging spectrometers based on the Bomem MB series of FTIR spectrometers The imaging spectrometers DREV ISR 0 IISR I and ISR II where developed for the Defense Research Establishment at Valcartier DREV Canada These imaging spectrometers are especially interesting for FOA since they are based on the same type of spectrometer as the one used in the ScanSpec system The principle of the design of those spectrometers is rather simple Instead of using only one detector to record one interferogram see Figure 5 a 2D array of R detectors is used Each detector will then record interferograms from different spatial locations The arrangement can be seen as many independent Fourier transform spectrometers that use the same optics The development of the first Infrared Imaging Spectral Radiometer ISR based on a Bomem FTIR spectrometer started in 1988 It was based on the largest InSb detector array commercially available at the time 4x8 and is called DREV ISR 0 It was a bench top proof of concept instrument and it was followed by DREV IISR I that
67. om FOA Daedal does not have an off the shelf product that matches our needs One possible way would be to buy separate parts from Daedal but for the alternative with separate parts Newport had better products to match our specifications Other alternatives Products from other companies have also been studied Aerotech internet Aerotech offers a complete gimbal system shown in Figure 15 The weight of more than 100 kg was way too much and the price is probably not even close of matching the budget Figure 15 Gimbal system from Aerotech Other companies that have similar products as the ones above include Ball Aerospace amp Technologies Corp internet Ball and Anorad internet Anorad Alternative Newport the chosen alternative Newport could offer a complete gimbal system for our needs but it would have to be custom built and the price would be closer to 200 000 DEM than the required maximum cost of 200 000 SEK However Newport sells separate parts that suit our application They sell both rotary tables and cradles Some of their products are shown in Figure 16 31 RIM 20 P nS ee Figure 16 Rotary tables left and cradles right from Newport The rotary tables have better accuracy then the cradles Therefore two rotary tables were bought One rotary table plus a cradle was also considered as an alternative The two rotary tables RV120PP and RV160PP were bought with stepper motors controlled by the moti
68. on controller MM4005 From now on the motion controller MM4005 will be called MM Figure 17 A photo of the mirror turning system of ScanSpec One of the two power inhibit switches is marked The other parts of the mirror turning system are labeled in Figure 18 How the two rotary tables are mounted can be seen in Figure 17 and Figure 18 The mirror turning system have two emergency motor power inhibit switches one shown in Figure 17 that protect the tables from rotating too far which would cause seriouse damage to the construction The use of AutoCAD AutoCAD was used when making the construction of the mirror turning system and the rest of the ScanSpec hardware The dimensions of the rotary tables from Newport were known in advance so by using CAD Computer Aided Design everything could be constructed before the rotary tables arrived This was almost a prerequisite to be able to finish the project on time Making this design 32 using only paper and pen instead of by modeling of solids would have been very much harder Another advantage with AutoCAD is the the program has built in functions to compute the moments of inertia which were needed in order to check the torque needed from the rotary tables for a certain angular acceleration mirror mounting point RV160PP RV120PP mirror mounting points U beams Figure 18 AutoCAD model of the mirror turning system The mirror turning system of the AutoCAD model is shown in
69. pec software 7 1 General The software for ScanSpec is based on the Matlab programming language because it is easy to program It has good tools for graphical presentation of data and the staff at the department of IR systems at FOA know the language and will be able to make modifications in the code for the ScanSpec software However the C and C programming languages have been used for communicating with external hardware Matlab does not currently have the ability to communicate with a serial port of a computer for example C has also been used once when the corresponding Matlab code executed much too slowly A special motion control programming language from Newport has been used to control the two rotary tables A few parts of the Matlab code rely on the Matlab Signal Processing and Image Processing toolboxes The details of the implementation will not be covered here Please refer to the comments in the code and to the Software manual filename Software manual txt provided with the ScanSpec software for detailed information The first release of the ScanSpec software version 98 1 is available on a CD complete with documentation CD ScanSpec98 1 There is also a FOA internal report manuals which contains a print out of the Software manual and the User s manual for ScanSpec The software consists of three modules briefly described below Module 1 Acquiring data This module lets the user choose a scan program and start a
70. quired When the acquisition of the interferograms is finished the data is sent from the acquisition board to the RAM memory of the FTIR computer With Acquire the user can then produce calibrated spectra from the acquired data and calibration files The calibrated data is stored in a special file format that can be read by the ScanSpec software Unfortunately the Acquire program cannot be controlled from another program The user has to interact with Acquire to produce the calibrated spectra from the acquired data so this process cannot be automated If the resolution 16 cm is chosen a maximum of 65 spectra can be acquired per second For the resolution 4 cm the spectrometer can acquire 31 spectra second The spectrometer can add many interferograms to reduce noise before the data is sent to the acquisition board in the FTIR computer The field of view of the telescope is the angle of the circular cone from which the telescope gathers light It is abbreviated FOV 19 Each spectrum gets a time stamp when it is acquired based on the system clock of the FTIR computer It is not specified in the documentation from Bomem Inc Bomem1 Bomem2 if the time stamps mark the beginning the middle or the end of the time period while the spectrum is being acquired 0 13 0 12 0 11 0 1 time s 0 09 0 08 0 07 WLLL 0 150 200 250 300 350 400 sample number Figure 7 The time between two consecutive spectra 21 o
71. r as ASCII text which increases the amount of data many times compared to sending it as binary data The number of samples of the positions of the rotary tables should not be too high Approximately 6 positions of both the rotary tables are sampled per scan line The spatial oversampling parameter mentioned in section 7 1 demands further explaination Generally the errors of the measured spectra from a Fourier transform spectrometer can be significantly reduced by averaging several interferograms for each spectrum to reduce the random noise This is also the case for the Bomem MR254 spectrometer This can be done by choosing to average several interferograms in the Acquire program But if only one spectrum is acquired per pixel field of view the image will be smeared out because the mirror is moving continouosly during the acquisition of the spectrum causing motion blur To handle this another way to reduce the noise has been implemented If the spatial oversampling parameter is set to N N spectra are acquired per pixel in the spectral image How the noise will be reduced by this method will be described Module 1 creates the following data e The sampled positions from the MM e The file s current raa and or current rab Current raa contains the acquired spectra from detector A and current rab the spectra from detector B It also contains a time for each spectrum e Some parameters that describe the size of the image the spectral resolution t
72. r is inexpensive and the delivery time short A mirror according to Alternative 2 could then be bought The mirror turning system is designed for the Alternative 3 mirror but it also works for the Alternative 2 mirror even though it will then be a bit slower The mirror from Edmund Scientific was tested with the telescope and the CCD camera on the spectrometer The results showed that the flatness of the mirror is good enough The image recorded by the CCD camera was not perfectly sharp not as sharp as when the 10 mm float glass mirror was tested but not so blurred that it matters much for this application 6 2 The mirror mounting When choosing the way to mount the mirror I considered essentially five requirements 1 The mirror mounting should not introduce unnecessary forces on the mirror which causing the mirror to bend 2 The mounting should be so stiff that the angles between the surface of the mirror and the mirror turning system do not change more then 0 01 during the rotations 3 The mounting should be made so that bending due to angular accelerations and the mass of the mirror is minimized 4 The strength of the mounting should be enough to carry the weight of the mirror and other forces 5 Thermal expansion should be considered It is almost impossible to try to support the mirror with something flatter than the mirror itself So the mirror has to support its own weight and it must also handle the forces due to angular ac
73. rared electromagnetic radiation X gt Figure 2 This figure depicts how a spectral image can be seen as a collection of several gray scale images one for each frequency for which the spectral radiance has been recorded An alternative way of mentally visualizing the data of a spectral image is as a collection of many ordinary gray scale images one for each frequency for which the spectral radiance was recorded This way of visualizing a spectral image is depicted in Figure 2 and has been the basis for the presentation of the data collected by ScanSpec From now on we will concentrate on spectral images in the infrared part of the electromagnetic spectrum which can be divided into the following bands NIR near infraread 0 7 1 1 um SWIR short wavelength infrared 1 1 2 5 um MWIR mid wavelength infrared 2 5 7 um LWIR long wavelength infrared 7 15 um FIR far infrared gt 15 um The definition of these frequency bands differ in the literature I have followed the definition by Holst ScanSpec covers the whole region from NIR to FIR but it is less sensitive in the NIR and the FIR region 2 2 The history of spectral images The astronomer George Ellery Hale is credited with being the first person to create a spectral image At the turn of this century he realized how important it was to study both the spectral and the spatial variations of the light from the sun He designed a spectrometer that would only let through light at
74. rature in the room was about 20 C and the temperature around the edges of the aperture of the blackbody source was not very much higher The temperature of the blackbody did not change more than 3 K during the same measurement According to Wien s radiation law the total radiation emitted by a blackbody is proportional to T This means that a temperature change of 3 K will change the measured values with approximately 3 4 1200 273 which is less than 1 per cent 52 Appendix 1 The spectrometer was set at alignment mode with a spectral resolution of 4 cm This way the total radiation could be measured The gain for the detectors was never changed during a measurement Figure 1 shows the equipment that was used The angle of interest Ad can be approximated with arctan hXd a to a very high accuracy since h lt lt d h lt 0 20 m and d gt 35 m ie cardboard m SS grid h V blackbody a source pan axis of the aperture of the black tripod body source Figure 1 The equipment for the measurements seen from above The distance a is the distance from the pan axis of the tripod to the primary optics of the telescope not to the front end of the telescope a was measured to be 0 55 m Measurement 1 The focusing range The CCD camera was used to measure the shortest distance an object can be focused at by the telescope The result was 3741 m from the front end of the telescope According to Bomem 1994
75. rough the surface of the mirror The advantage of this construction is that the system would cover a large part of the sky and that the target following feature would be provided for The disadvantages are that two mirrors must be used and that the supporting construction for the mirrors and the spectrometer would be more complicated than for the alternatives below that use only one mirror The target following feature could also be realised with only one mirror Alternative 2 as shown in Figure 9 spectrometer up telescope 90 90 mirror Figure 9 Main construction Alternative 2 The elevation angle can be between 4 and 50 and the azimuthal angle from 90 to 90 from straight forward Alternative 2 has the advantage that it only uses one mirror which increases the performance On the other hand the standing arrangement is harder to stabilize to the ground since the center of gravity of such a system would be much higher above the ground than for Alternative 1 There is also another problem The spectrometer itself works well in a vertical angle as in Figure 9 but the IR detectors are cooled with liquid nitrogen and the dewars for the nitrogen are not made for vertical use of the 21 spectrometer It is possible but costly to make new dewars designed specifically for the vertical use of the spectrometer Paul Coulson at Bomem Inc guessed the cost for such dewars to be in the order of 3
76. rsvarets forskningsanstalt i Link ping Sverige Syftet med projektet var att bygga en bildalstrande spektrometer Detta har gjorts genom att montera ett l ngt teleskop p en snabb FTIR spektrometer och genom att rotera en spegel framf r teleskopet P detta s tt samlas spektra fr olika riktningar in Spegeln anv nds till att skanna av en scen f r att samla in en spektral bild det vill s ga en bild d r varje pixel best r av ett helt spektrum Denna rapport innefattar en versikt av bildalstrande spektrometrar f r IR omr det och av den milit ra anv ndingen av spektrala bilder Utvecklingen av den bildastrande spektrometern ScanSpec beskrivs och de f rsta m tningarna presenteras Rapporten r skriven f r den som r intresserad av spektrala bilder bildalstrande spektrometrar och speciellt den bildalstrande spektrometern ScanSpec L saren antas ha en del allm n fysikkunskap p universitetsniv Nyckelord spektrala bilder bildalstrande spektrometer FTIR ScanSpec Ovriga bibliografiska uppgifter Spr k Engelska amerikansk ISSN 1104 9154 Omf ng 64 s Pris Enligt prislista Distribut r om annan n ovan V 1 1 Contents De BLM OCUCEIO NN 2o5cce si cinecesecescceeeseececececesdeeesussseensesseeessussseusvensseuvesssseuscussseussseauseceussdeusesesstocsessses 6 PART I SPECTRAL IMAGES 2 Spectral images and their military USE sscccssssssscssssscscsscssccsssssscsssscsscsssscssessssssseeees
77. s an upper limit of the spatial frequencies that can be measured with the scanning system The telescope works as a low pass filter for the spatial frequencies Appendix 1 contains figures showing the magnitude of the Fourier transform of the angular sensitivity functions the MTF s for the two detectors together with the narrow angle telescope Note that these measurements were performed The magnitude of the Fourier transform of the angular sensitivity is commonly called the MTF the modulation transfer function See Holst for a thorough discussion of MTF 23 without the mirror that ScanSpec uses The mirror also affects the angular sensitivity functions The figures of the Fourier transforms in Appendix 1 show that spatial frequencies above 5 cycles degree will be almost totally destroyed by the low pass filtering feature of the telescope Thus spatial frequencies above 5 cycles degree are not possible to acquire According to the Nyquist criterion a sampling frequency of 10 cycles degree is enough to reproduce a signal that is limited to frequencies below 5 cycles degree This sampling frequency is used with ScanSpec that is a spectrum is acquired every 0 1 One figure of merit when quantifying the amount of aliasing is to see how much of the total power of the MTFs that falls within the frequency band that can be reconstructed within 5 cycles degree in this case Watson Crude estimates show that that figure is above 88 for th
78. s much more tolerant of timing errors then the scan pattern shown in Figure 11 The timing problems will be further discussed in the software chapter chapter 7 3 25 6 The ScanSpec hardware 6 1 The mirror When choosing a mirror for IR applications there are some important parameters of the mirror to consider reflectance flatness protective coating size and shape weight and price Reflectance The reflectance of a mirror that will be used for IR applications is important in two ways Firstly a high reflection is needed to keep most of the radiation that will be studied Secondly and often more important a high reflection is needed to decrease the amount of thermal emission from the mirror This emission of R radiation from the mirror is especially harmful it is hard to calibrate the spectrometer so that this radiation will be compensated for since it might not be possible to keep the temperature of the mirror constant Gold coated mirrors have best reflectance in the R region of the spectrum Silver and aluminum coated mirrors also have good reflectance in this region and they have higher reflectance than gold in the visible region Flatness The flatness of the mirror is not very crucial in this case The mirror will only collect light for one pixel of the image at a time The flatness is much more important for an imaging system that collects a whole image at a time One problem when buying a mirror is that different ma
79. s to let the mirror go to the desired starting position and stop Then the spectrometer gathers one spectrum and the mirror moves to a position corresponding to the next pixel in the spectral image and stops Then the spectrometer would gather the next spectrum and so on This way of step scanning would require enormous accelerations of the mirror if the scanning should be so fast that the speed of the spectrometer limits the time of acquiring a spectral image Another problem with this type of scanning is that the spectrometer cannot be synchronized with the step motion The other way of scanning the mirror which was chosen in this project is to continously scan the mirror over one line of the image at a time By scanning all the lines all the data for the spectral image can be acquired The drawback with continous scanning is the motion blur due to the fact that the mirror is moving during the acquisition of the spectra Note that a Fourier transform spectrometer gathers data for all the frequencies at the same time the Fellgett or multiplex advantage Pedrotti so that all frequencies will be recorded from the same field of view The motion blur is not such a serious problem since the mirror will never move more than one pixel distance during the acquisition of one spectrum Also since the angular speed of the mirror is known the motion blur could be compensated for in the signal processing of the data However this has not been done since motion
80. sampling 1 2 scan per spectrum Standard scan program 2 NORMAL 16 16cm 16x16 4x spatial oversampling 1 2 scan per spectrum Standard scan program 3 NORMAL 32 16cm 32x32 no spatial oversampling 2 scans per spectrum Standard scan program 4 FAST 8 16cm 8x8 no spatial oversampling 1 2 scan per spectrum Standard scan program 5 FAST 16 16cm 16x16 no spatial oversampling 1 2 scan per spectrum Standard scan program 6 FAST 32 16cm 32x32 no spatial oversampling 1 2 scan per spectrum Standard scan program 7 LOW NOISE 8 16cm 16x16 2x spatial oversampling 16 scans per spectrum Standard scan program 8 LOW NOISE 16 16cm 16x16 2x spatial oversampling 16 scans per spectrum Standard scan program 9 LOW NOISE 32 16cm 32x32 no spatial oversampling 16 scans per spectrum Standard scan program 10 HIGH RES AND LOW NOISE 4cm 8x8 2x spatial oversampling 16 scans per spectrum Standard scan program 11 HIGH RES 16 4cm 16x16 4x spatial oversampling 1 2 scan per spectrum Furthermore these settings are used with all standard scan programs scan_mode 1 use_A_detector 1 use_B_detector 1 oversampling_mode_A 0 oversampling_mode_B 0 aperture 1 pixels_per_FOV 2 telescope 1 See a standard program file in CD ScanSpec98 1 for exact definitions of these parameters 64
81. spectra on the scan line was acquired by the spectrometer If the position is seen as a function of time this can be done by linear interpolation and extrapolation using the positions recorded by the MM Linear interpolation and extrapolation is well justified in this case since the speed of the motion is constant while the line of sight is on a scan line An example of what the positions of the spectra can look like is seen in Figure 23 The positions of the spectra which were not acquired on a scan line are also marked even though they are never estimated in the signal processing A spatial oversampling of 2 has been used in the example in Figure 23 Normally there are two samples per pixel but not always due to the irregularity of the times between the spectra as exemplified in Figure 7 page 20 38 Part Il Computation of the spectra in the spectral image From part I of the signal processing we have the positions of all the spectra acquired on a scan line These positions will generally not coincide with the positions of the center of the pixels in the spectral image We want to estimate the spectra in those points Let us consider the spectral radiance of one specific scan line and frequency That spectral radiance will then be a function of only 02 as depicted with a stem plot in Figure 24 A Note that at two places one spectral radiance value is missing and at another place two radiance values are missing This is because of the irregularity o
82. t the moment when the trigger signal is sent to the parallel port It might take a few hundredths of a second before the processor starts executing the C program again Maybe the problem could be solved by using the interrupts of the Pentium processor 40 The scan pattern was chosen so that the system can tolerate some timing error The slow standard scan programs Program 1 2 7 8 9 10 and 11 see Appendix 3 are not affected by this timing problem If the timing error is too large for the fast scanning programs Program 3 4 5 and 6 the data for the spectral image can be recalculated with a time correction given by the user 41 8 The first measurements All the data from the measurements described below are available on the CD CD ScanSpec98 1 in the directory ScanSpec data 8 1 Measurements with the standard scan programs The very first measurements made with ScanSpec were performed on a collimator target The collimator at the Department of IR systems at FOA can be used to simulate a target at an infinite distance and with a certain temperature and angular extent All the eleven scan programs were tested on the same target One example of the result for a four bar target is shown in Figure 25 page 44 It is made with scan program 6 that acquires a 32x32 spectral image with a spectral resolution of 16 cm in about 30 seconds No spatial oversampling or averaging of interferograms was used The temperature difference between t

Download Pdf Manuals

Related Search

Related Contents

MICRO TECH ŒŒIÏIIJBŒ III  Manuale Utente O-Ring  Baixar manual  AmiVoice ® Ex7 簡易操作マニュアル(使用開始時)  Etiquette Produit  Omron M7 User's Manual  T'nB CS10 headphone  設定ガイド    Ewent EW3128  

Copyright © All rights reserved. Failed to retrieve file